AI LAB
Kettl & Co. Support Assistant: Grounded RAG with Citations
Build a project on grounded RAG that answers support questions from product manuals with citations
Verified Certificate on Successful Completion
What is in it for you?
LLM tokens included
A metered token budget is bundled in, call real models from your code with no API key or extra cost.
AI mentor on tap
Stuck? An in-IDE AI mentor gives hints and debugging help, without handing you the answer.
Instant grading
Submit and get an objective, rubric-based verdict in seconds, pass, or actionable feedback.
Verified certificate
Pass and earn a shareable, verifiable certificate you can add to your LinkedIn profile.
Zero-setup cloud IDE
A ready sandbox with the libraries pre-installed, start building in the browser instantly.
Retry until you pass
Iterate as many times as your budget allows; a fail keeps the lab open to try again.
Grab your slot before the offer expires
Start solving today!
Basic Info
Select Offers
Project Statement
Business Problem
Kettl & Co.'s support team of 25 agents fields roughly 4,000 tickets a month, and 60% of them ask questions already answered in the product manuals. Agents burn an average of 6 minutes per ticket hunting through docs and re-typing the same answers, and generic chatbots make it worse by confidently inventing warranty terms and specs that don't exist.
AI Solution
You'll build a retrieval-augmented assistant that chunks the manuals, embeds them into an in-memory vector index, and retrieves the most relevant passages for each customer question. The assistant generates answers grounded strictly in the retrieved text, attaches a citation to the source passage, and honestly says 'I don't know' when the answer isn't in the docs.
Potential Impact
Grounded, cited answers can cut per-ticket handle time from 6 minutes to under 2, freeing an estimated 250+ agent-hours per month for complex cases. Eliminating fabricated answers protects CSAT and cuts costly warranty disputes, while citations give agents instant, trustworthy source verification.
Skills that you will build
Solve a real AI-engineering problem that builds in-demand skills and a portfolio-ready, verified certificate.
.ffb97545.png&w=16&q=75&dpl=dpl_8zM6HG9D1ZAbdMYa1sGJ7PjNFzPo){kind=small}
Document chunking strategy and tuningGenerating and using text embeddingsBuilding an in-memory vector index with FAISSSemantic retrieval of relevant passagesGrounding LLM answers in retrieved contextAdding source citations to generated answersHandling out-of-scope questions to prevent hallucinationAI Lab Pre-requisites
- Comfortable writing Python functions
- Basic understanding of how LLMs generate text
- Familiarity with what embeddings are (helpful, not required)
Target Roles
- AI Engineer
- LLM Application Developer
- RAG Systems Engineer
- AI Solutions Architect
Combo Offers
Most Frequently Bought
🛡 Verified certificate on pass
Super Saver Offer
🛡 Verified certificate on pass
Create Your Own Combo
🛡 Verified certificate on pass
for successfully completing the 'Kettl & Co. Support Assistant: Grounded RAG with Citations' project
Add a Industry Recognized
Project Certificate To Your Resume
Industry Recognized
Project Certificate
Career Advancements
Elevate your career with a respected certificate
Industry Respect
Gain credibility in the field
Networking
Connect with experts and peers
Opportunities
Attract exciting job prospects and promotions
for successfully completing the 'Kettl & Co. Support Assistant: Grounded RAG with Citations' project
Frequently asked questions
Are LLM tokens included?
Yes. The lab runs in a cloud IDE with a metered LLM budget included — no separate API keys or costs needed.
What do I earn on completion?
A verified, shareable credential you can add to your resume and LinkedIn to prove you can build grounded RAG systems.
Do I need prior RAG experience?
No. If you can write Python functions and understand basic LLM concepts, this lab guides you through building retrieval, grounding, and citations from scratch.
Will an AI mentor help me if I get stuck?
Yes. An AI mentor is available in the IDE to nudge you toward the right approach without handing you the solution.
How is my work evaluated?
Your assistant is auto-graded on whether it retrieves and grounds answers correctly, cites sources, and refuses to fabricate answers for out-of-scope questions.
How long does this lab take?
Most learners finish in about 90 minutes, depending on your chunking and retrieval experimentation.